基于Kalman滤波与神经网络的高精度同步时钟算法

大规模分布式电源、储能与电动汽车的接入对配电网状态监测与运行控制带来了挑战。基于配电网同步相量测量单元(phasor measurement unit,PMU)的广域量测系统被认为是解决这一问题的有效方式。然而,采用低成本晶振时,现有的同步时钟算法难以满足配电网PMU对同步时钟高精度、高稳定性、低成本的要求。为满足配电网PMU应用的需求,提出一种基于Kalman滤波器与BP神经网络的授时/守时算法。基于卫星信号误差与晶振频率数学模型,利用Kalman滤波器对卫星信号的随机误差进行滤除,提高授时精度,并提供准确的晶振状态数据。利用此数据训练BP神经网络模型,刻画出晶振频率的老化规律,提高守时性能。在卫星信号正常接入与失锁场景下,基于实际时钟装置量测数据进行测试验证。测试结果显示,文中所提算法在不提高现有硬件成本的基础上,有效提高了同步时钟的算法性能。

A High Accuracy Synchronous Clock Algorithm Based on Kalman Filter and Neural Network

Access of large-scale distributed power supplies, energy storage and electric vehicles brings challenges to state monitoring and operation control of distribution network. The wide area measurement system based on synchronized phasor measurement unit(PMU) of distribution network is considered as an effective way to solve the problem. However, existing synchronization clock algorithm using low-cost crystal oscillator can hardly meet the requirements of high accuracy, high stability and low cost for the PMU of distribution network. In order to meet the demand of PMU application in distribution network, this paper proposes a timing/time keeping algorithm based on Kalman filter and BP neural network. Based on the mathematical models of satellite signal error and crystal oscillator frequency, Kalman filter is used to eliminate the error of satellite signal to improve timing accuracy and provide accurate crystal oscillator state data. The data are used to train a BP neural network model, describing the aging rule of crystal vibration frequency and improving the time keeping performance. In the scenario of normal access and losing lock of satellite signal, the tests are carried out based on the measured data of actual clock device. Results show that the algorithm proposed in this paper can effectively improve the performance of synchronous clock without increasing cost of existing hardware.